AI may aid screening for opioid use disorder

An AI screening tool might help to spot people who are at risk for opioid use disorder.Westend61 on Offset / Shutterstock

Opioid use disorder (OUD) remains a serious problem in the U.S. People with OUD who are hospitalized are at increased risk for overdose, hospital readmission, and other complications. Hospital-based addiction treatment has been shown to improve outcomes and reduce death rates. However, hospital-based screening for OUD can be inconsistent. Many at-risk patients are released before seeing an addiction specialist—a situation that’s been linked to increased overdose rates.

An NIH-supported research team led by Dr. Majid Afshar of the University of Wisconsin-Madison aimed to see if AI-based tools might help. They previously developed an AI screening method to detect hospitalized patients at risk for OUD. Their screener can rapidly analyze data in patients’ electronic health records and alert health care providers if it finds patterns consistent with OUD. The AI system may then recommend consultation with addiction specialists and other interventions.

In their new study, the researchers tested the AI-based screener to see if it could lead at-risk patients to meet with an addiction specialist. The team also looked at re-hospitalization rates 30 days after discharge. The clinical trial included more than 51,000 adults admitted to the University of Wisconsin Hospital. Results were reported in Nature Medicine on April 3, 2025. 

During an initial baseline phase—from March to October 2021 and March to October 2022—the team gathered data on provider assessments of patients’ OUD risk and outcomes. This phase included about 34,000 patients. The AI screener was then assessed in a second phase, from March to October 2023. In this phase, health care providers used the tool to help assess more than 17,000 patients. A total of 727 addiction medicine consultations were completed during the study period.

Overall, the AI screener was as effective as provider-only assessments in leading to addiction specialist consultations. However, those who received AI screening were 47% less likely to be readmitted to the hospital within 30 days after initial discharge.

The team calculated that each readmission avoided saved about $6,800 in health care costs during the study period. The findings suggest that investment in AI could help to increase access to addiction treatment, improve efficiencies, and save costs.

“AI holds promise in medical settings, but many AI-based screening models have remained in the development phase, without integration into real-world settings,” Afshar says. “Our study represents one of the first demonstrations of an AI screening tool embedded into addiction medicine and hospital workflows, highlighting the pragmatism and real-world promise of this approach.”